CV

Senior Applied Scientist

Amazon Ads & Special Projects

• Tech Lead & Single Threaded Owner — TACTIC: Designed and launched the first agentic targeting clause promotion/demotion system for 69M+ worldwide ad campaigns; estimated $3M+ in ads revenue impact. 2+ patent applications in progress.
• Tech Lead & Single Threaded Owner — AutoTrends: Autonomous trend-aware LLM agentic system using a two-stage gated trend engine for detecting implicit and explicit trends in targeting; 1.3× performance lift across 69M+ campaigns.
• Single Threaded Owner — Distributed Optimization System: Scalable DSPy-powered explainability framework orchestrated via Apache Airflow on EMR clusters; beta launched April 2026.
• Mentored junior scientists on agentic AI architectures and CoT evaluation frameworks, establishing team-wide best practices.

Principal Scientist

SAP Labs

• Shipped SAP Consultant Capability (SCC) to General Availability (Q4 2024) — flagship enterprise RAG product driving $22M in business impact, serving 22K multi-turn conversations/day at precision 0.79.
• Formally managed cross-functional team; grew direct reports from 2 to 8 (including interns); scaled total team to 16. Conducted structured 1:1s, set OKRs, delivered performance reviews.
• Architected end-to-end KG-RAG pipeline; fine-tuned BGE-EN-ICL retrieval model and Claude 3.5 Sonnet generation model; deployed multi-node Milvus DB; optimized inference via vLLM.
• Developed strong RAG evaluation framework incorporating domain experts and RAGAS-based LLM-as-a-judge methodology (patent application in progress).
• Published at ECIR 2025 and SIGIR; 3+ patent applications in progress.

Senior Research Scientist

Dataminr

• Launched RAG-based fact retrieval system using E5-Mistral embeddings and fine-tuned LLaMA-7B; achieved 82% factual accuracy on production data processing 19M tweets/day.
• Designed hallucination detection and correction system for summarization pipeline (mBART + GPT), correcting 802K hallucinated entities per day at 66% accuracy.
• Built end-to-end knowledge graph construction pipeline (Neo4j) processing 19M tweets/day; fine-tuned BERT for NER, LUKE for relation extraction (precision 70.2); orchestrated via Apache Airflow and Databricks.

Machine Learning Research Manager

Verisk AI

• Formally managed a team of 5 MLEs — conducted hiring, structured 1:1s, set quarterly OKRs, and delivered performance reviews.
• Secured $250,000 in DARPA funding by assembling a team and leading proposal submissions — a net-new revenue stream for Verisk.
• Shipped abstractive summarization system for large-scale multimodal event news (ROUGE-F1 > 35%, deployed on Apache Airflow).
• Shipped event knowledge graph (‘The Machine’) with 1M+ entities and 80 property edges, driving $1M+ business need for Verisk clients.

Acting Senior Key Expert / Staff Research Scientist

Siemens Corporate Technology

• Co-Principal Investigator on $1.8M DARPA ASKE program (Deep Code Curator); led 15 members across industry and academic partners across two successful program phases.
• Managed 2 research scientists and 2 software engineers on multimodal KG and visual QA projects.
• Built end-to-end BiLSTM-CRF extraction pipeline converting 960 scientific papers into queryable RDF knowledge graphs.
• Developed KG-enabled caption generation and Visual QA systems using DBpedia/YAGO, encoder-decoder networks, trained on COCO/Pascal VOC datasets.
• Promoted from Staff Research Scientist to Acting Senior Key Expert within one year based on DARPA program leadership impact.

Senior Systems Engineer

Infosys Technologies

• Built iSEE (Infosys Sentiment Extraction Engine) for customer opinion mining; resulted in a granted US patent (2013).
• Developed Semantic Wiki integrations and BPEL-based workflow automation for enterprise clients; published at WWW 2009.

PhD

Rensselaer Polytechnic Institute

Computer Science

• AAAI 2016 Doctoral Consortium Award.
• 3MT Thesis finalist.

M.S.

Rensselaer Polytechnic Institute

Computer Science

B.E.

Anna University

Computer Science

Co-PI, DARPA ASKE Program — $1.8M Award

Defense Advanced Research Projects Agency (DARPA)

Co-Principal Investigator on the $1.8M DARPA ASKE (Automating Scientific Knowledge Extraction) program at Siemens Corporate Technology.

AAAI SIGAI Doctoral Consortium Award

Association for the Advancement of Artificial Intelligence

Selected for the AAAI 2016 Doctoral Consortium, recognizing outstanding doctoral research in AI.

NSF Award

U.S. Semantic Technologies Symposium (US2TS 2018)

NSF travel and recognition award at the U.S. Semantic Technologies Symposium.

Three Minute Thesis (3MT) Finalist

RPI Graduate Research Symposium

Finalist at the RPI Graduate Research Symposium 3MT competition.

Efficient Knowledge Graph Construction and Retrieval from Unstructured Text for Large-Scale RAG Systems

arXiv

Towards Practical GraphRAG: Efficient Knowledge Graph Construction and Hybrid Retrieval at Scale

arXiv

Combining Knowledge Graphs and Retrieval Augmented Generation for Enterprise Resource Planning

ECIR 2025

Multimodal Knowledge Graph for Deep Learning Papers and Code

CIKM 2020

Diag2Graph: Representing Deep Learning Diagrams in Research Papers as Knowledge Graphs

ICIP 2020

Deep Code Curation (DCC) — A Project to Build Multimodal Knowledge Graphs from Deep Learning Papers

AKBC/SLKB Workshop 2019

Feature-based Reformulation of Entities in Triple Pattern Queries

arXiv

In-context Query Reformulation for Failing SPARQL Queries

SPIE 2017

Leveraging Semantics for Large-Scale Knowledge Graph Evaluation

WebSci 2017

Pragmatic Querying in Heterogeneous Knowledge Graphs

AAAI 2016 Doctoral Consortium

Open Government Data: A Data Analytics Approach

IEEE Intelligent Systems

Semantic Wiki Aided Business Process Specification

WWW 2009